Intelligent Coastal Surveillance Automat for Object Detection

odecrackAI and Robotics

Oct 29, 2013 (3 years and 10 months ago)

66 views

Intelligent Coastal Surveillance
Automat for
Object Detection



S. Itzikowitz
, S
. Sheraizin
,

Intelligent Video Systems R&D Institute, Computer Science Department,

Academic Studies Division, College of Management

E
-
Mail:
samitz@st.colman.ac.il

,
semion@cs.colman.ac.il


Abstract


A coastal surveillance system
operates
in a predefined area. It is intended
to detect
a
suspicious object that
appears

with
in
this

area.

A
utomatic object detection in
a
coastal surveillance system leads to some

specific
problems
since

a coastal environment

is a random changing background
caused
by water
waves and light

changes
. Moreover
,

there are spatial and temporal changes characterized
by
a
large dynamic range.

The suspicious objects are humans and
surface
vessels/ships
.

H
umans
enter

the
observation area
either
from
the
coastline

or
from the
water (swimm
ers
).


We
have
developed
an

intelligent coastal surveillance automat
,

iCSA
,

based on

a
multistage robust object detection method and

algorithm providing a low false detection
probability. The method takes

into account a multisensor use with either
both
the
luminance and chrominance data or the luminance data only.
Each

stage improves
the

o
bject detection
by
using
a
set of high
ly

informative descriptors.
Certainly,

object

detection in
a
complicated environment requires complicated image data preparation
before
object

detection,
as well as
accumulation of some data analysis

partial results an
d
decoding
of
the multistage object detection results.


The iCSA image data processing method provides the following unique

features:


* Low false detection, recognition and tracking by using the full


automatic multistage image data

processing.


* Work possibility in dark area, wide range of a light change,




intensive water waves, wind and rain.


* Effective detection, recognition and tracking of a person

or persons




entering

the predefined
area from

the
water or
from
the
coastline
,




and various

types of ships.




Experimental Results


Our method, algorithm and SW

were tested

on Korean video clips

with the following
result
:




*
Our

method, algorithm and
softwar
e provide

full automatic detection,
recognition and tracking of any intruder moving
along the

coas
t
line
or
in
the
water

as well as
suspicious movements of

ships and surface
vessels
.



Conclusions
:




*
Our

method, algorithm and SW allow full automatic detection,
recognition and tracking
of
a

person or persons
enter
ing
an
observation
area
from
the
coastline

or from
the
water side, 24/
7.


*
Our

method,

algorithm and SW allow full automatic detection,
recognition and tracking
of a ship

entering
an observation area.


*
Our
method and algorithms can be used as

basic solutions for

an
Intelligent Coastal Surveillance





System Automat
(
iCSA
)

development
,

providing full
automatic object (
person, ship, etc.) detection,


recognition and tracking.